plankton-code-quality
Pass
Audited by Gen Agent Trust Hub on Mar 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill configures PreToolUse and PostToolUse hooks that automatically execute shell scripts (multi_linter.sh, protect_linter_configs.sh, stop_config_guardian.sh) on every file modification within the agent's environment.- [COMMAND_EXECUTION]: It performs automated remediation by spawning Claude subprocesses (claude -p) which are granted the authority to modify source code based on linter output.- [COMMAND_EXECUTION]: The skill enforces specific package managers (uv, bun) by intercepting and blocking standard bash commands for pip, npm, yarn, and poetry.- [EXTERNAL_DOWNLOADS]: Directs the user to install a large number of third-party tools from various package registries (Homebrew, PyPI, NPM), including jaq, ruff, uv, biome, shfmt, taplo, markdownlint, shellcheck, yamllint, hadolint, oxlint, semgrep, and knip.- [DATA_EXFILTRATION]: Automated remediation involves sending project code and linter violation data to external model providers for processing.- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection where untrusted data in the codebase (such as comments or strings) could influence the linter output, which is then processed by an automated LLM subprocess with write access to the filesystem.
- Ingestion points: Linter violation summaries (JSON) in SKILL.md.
- Boundary markers: None specified for the data passed to the remediation subprocess.
- Capability inventory: The skill spawns claude -p subprocesses that modify files in the repository.
- Sanitization: No sanitization or validation of linter-generated messages is mentioned before model interpolation.
Audit Metadata